Home   News   Article

Subscribe Now

Fiasco caused by algorithms




I have, from time to time, expressed suspicion about the significance and accuracy of computer models produced by ‘experts’.

The word algorithm has been prominent in discussion of the chaos surrounding this year’s ‘artificial’ exam results.

Algorithm is a word from computer science and is usually a complex set of rules that is thought to govern how a program or model will behave in arriving at an output. To an extent, model and algorithm are interchangeable for ordinary purposes.

In February, at the outset of the virus crisis, Government scientific advisers warned in clear terms that aggressive isolation policies would simply push viral transmission forward to the time when those policies were lifted but would not reduce infections overall.

This implied that isolation would need to be maintained indefinitely or until an effective vaccine was produced.

Imperial College then issued the notorious statistical modelling report, which panicked the government into imposing the lockdown.

Most other countries followed suit with disastrous economic consequences but were unable to stop viral infections, just as originally forecast.

Now all UK governments are embroiled in a fiasco over the exam results produced by ‘expert’ algorithms that do not agree with the ‘human’ forecasts produced by teachers who know their pupils.

The results awarded now have to be based on the direct experience of real experts in the real world and not on theoretical algorithms or models produced by remote ‘experts’.

If this is what happens, I believe that it may be a lesson for our times in considering any real-world mitigations of any present or future climate change rather than continue to rely on modelling projections of the future of climate.

One should remember the difficulty of forecasting the weather next week or next year never mind what it might be in 50 or 100 year’s time.­— ROBERT SHEPPARD, Hillside, Beckingham.



This site uses cookies. By continuing to browse the site you are agreeing to our use of cookies - Learn More